Chairman & CTO @OpenAI

Joined July 2010
Alice in Wonderland, as summarized by our latest AI system (openai.com/blog/summarizing-…):
24
64
6
431
Documentary on the OpenAI Dota project will be launched on twitch.tv/beyondthesummit, Friday, Sept 24th at 11a PT. Full-length screening followed by Q&A with the director, Dota community guests, and OpenAI team:
6
65
9
369
17,268
Cogram.ai—a Codex-powered tool that stops you from needing to constantly look up data science code (main reason I lose patience with data science myself!), letting you "focus on the problems that matter". Available in Jupyter Notebook. Demo: invidio.us/watch?v=bdnp0Ecc…
10
47
5
247
Race begins in T-1 hour — challenge.openai.com.
6
5
0
33
My personal favorite of our Codex examples — a program written in Python, which rewrites itself into Ruby, which rewrites itself into Python, which rewrites itself into Ruby, ad infinitum. Writing it felt like writing a new kind of quine (en.wikipedia.org/wiki/Quine_…).
11
30
3
190
11,061
Giving Codex a first grade math test:
17
176
33
895
47,481
The first GPT-3 NFT just sold for $478,000 at auction: thefirstinft.com/
16
57
22
229
We've found that it's possible to target GPT-3's behaviors to a chosen set of values, by carefully creating a small dataset of behavior that reflects those values. A step towards OpenAI users setting the values within the context of their application: openai.com/blog/improving-la…
14
35
1
182
A visualization of "pose" in the neuron that responds to the word "Celebration", in our latest neural net: openai.com/blog/multimodal-n… This network has neurons that respond to the same concept in images, text, and drawings — surprisingly reminiscent of multimodal neurons in humans.
6
22
2
195
Lazy Caturday:
7
3
0
139
.@theannabrockman you'll always be my Valentine.
3
5
1
139
Agreed:
Real progress in AI can only be achieved through a very intense work ethic
100
60
186
541
CLIP — our new neural network for classifying images into categories based on arbitrary text. Works like GPT-3's semantic search, but for images: andrewmayneblog.wordpress.co… Maybe a path towards vision nets that don't make silly mistakes: openai.com/blog/clip/
15
103
10
488
DALL-E — our new neural network for generating images from text: openai.com/blog/dall-e/
88
999
331
4,050
Hard to believe it's been five years already. Still feels like this to me:
Today is the five year anniversary of OpenAI. Super proud of what the team has accomplished in the first five years. Pretty sure the next five are going to be remarkable.
15
25
2
489
On this day, a couple of years ago, we went on our first date ❤️
16
5
1
343
Phil Brockman (my younger brother) just released his first Spotify album, in collaboration with @OpenAI MuseNet: open.spotify.com/album/7mget… Each song is 6-12 layered recordings of violin, which he played while having a metronome in his ear together with MuseNet's MIDI output.
11
17
1
178
Show this thread
Replying to @wear_here
We definitely omitted the "display" code, but the typical workflow for Q&A is: 1. Prepare a generic Q&A context 2. Concatenate in a question 3. Call the API to generate text 4. Return the newly generated text In the demo, the green is step 2; the output is the question + 4.
1
1
0
1